English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Modern cameras are not designed with computer vision or machine learning asthe target application. There is a need for a new class of vision sensors thatare privacy preserving by design, that do not leak private information andcollect only the information necessary for a target machine learning task. Inthis paper, we introduce key-nets, which are convolutional networks paired witha custom vision sensor which applies an optical/analog transform such that thekey-net can perform exact encrypted inference on this transformed image, butthe image is not interpretable by a human or any other key-net. We provide fivesufficient conditions for an optical transformation suitable for a key-net, andshow that generalized stochastic matrices (e.g. scale, bias and fractionalpixel shuffling) satisfy these conditions. We motivate the key-net by showingthat without it there is a utility/privacy tradeoff for a network fine-tuneddirectly on optically transformed images for face identification and objectdetection. Finally, we show that a key-net is equivalent to homomorphicencryption using a Hill cipher, with an upper bound on memory and runtime thatscales quadratically with a user specified privacy parameter. Therefore, thekey-net is the first practical, efficient and privacy preserving vision sensorbased on optical homomorphic encryption.

Key-Nets: Optical Transformation Convolutional Networks for Privacy Preserving Vision Sensors